Manang / Privacy

Privacy policy

Plain language. No surveillance, no telemetry, and complete transparency on what leaves your machine.

1. Your files stay on your disk

Manang is a local-first desktop application. Your papers, theses, slides, figures, notes, and bibliography databases live entirely on your own computer as normal files and directories.

We do not operate document servers, centralized databases, or cloud accounts. You do not need to create an account, log in, or stay connected to the internet to write, compile LaTeX or Typst, or draw vector figures.

2. Zero telemetry

The Manang application contains no analytics SDKs, no tracking beacons, no advertising libraries, and no silent phone-home pings.

  • We do not record what documents you open or how often you compile.
  • We do not record what packages, templates, or fonts you use.
  • We do not track keystrokes, editing duration, or error rates.
  • We do not collect IP addresses or fingerprint your hardware.

3. Third-party AI providers and API keys

AI features in Manang are strictly opt-in and require you to provide your own API key for an external model provider (such as Anthropic, OpenAI, or DeepSeek) or configure a local model (such as Ollama or llama.cpp).

Where your keys live:

Your API keys are stored securely inside your operating system's native credential store (such as the Linux secret service, macOS Keychain, or Windows Credential Manager) via manang-keystore. Keys never leave your machine except when sent as standard authorization headers directly to the API endpoint you configured.

What the AI provider sees:

When you invoke the assistant, Manang constructs a prompt containing only the explicit context required: your instruction, the selected text, or relevant project file fragments. Requests are sent directly from your computer to the AI provider. No prompt, document text, or API key ever routes through or touches Fundaments.work servers.

Your interaction with third-party AI endpoints is governed by the respective provider's terms and privacy policies. Commercial API endpoints typically commit not to train models on API inputs.

Global off switch:

You can inspect the exact payload before any request is sent, and you can disable AI features globally at any time in Settings.

4. Live co-editing (in beta)

Manang includes live collaborative editing, currently in beta. Real-time document updates are synchronized using end-to-end encryption.

The signaling and relay infrastructure routes encrypted messages between session participants. Relays only handle ciphertext and cannot inspect, decrypt, or log the text of your documents.

5. Diagnostic reports and feedback

When you select Help > Send feedback or report an issue, Manang generates a diagnostic summary in a local dialog. This report includes application version, operating system details, active plugin names, and recent sanitized log entries (with home paths and credentials automatically redacted).

Nothing is transmitted automatically. You inspect the text yourself and decide whether to copy and share it with us.

6. Contact

If you have questions about privacy in Manang, email us at contact@fundaments.work.